Federated Learning Usage of Brave for Ad Recommendations

I see that Brave is using federated learning and I saw that is a blog on brave’s website it was mentioned that Brave is using Federated Learning to predict for news prediction.

I was curious to know if Brave is using Federated Learning for ad-prediction and how is the matching between user’s watch history and advertisers requirement matching is done?

Can something like Private Set Intersection be used?

What is the input data on which brave does recommendations for the ad’s just browser’s history or is it with a combination of location data, mobile stats etc and forming some kind of cohort with the data available

Also do you’ll penalise for clickbaity data?